What it is
CollectUI is an image-based user interface (UI) inspiration gallery. Its own metadata says, “Discover amazing UI designs, portfolios, and inspiration curated by design magicians.” It sorts Dribbble Daily UI work and other Dribbble shots into categories such as cards, user profiles, portfolios, widgets, and admin panels.
It is not a design-token, DESIGN.md, or agent-spec product. It is a place for a person to browse screenshots and form a visual reference. That puts it in a different layer of the landscape from extractors and specification catalogs.
First principles
A screenshot and a design specification hold different information.
- A screenshot holds the result. Composition, density, hierarchy, and rhythm are visible at a glance. That is what a person needs for a mood board.
- A specification holds the decisions. It records exact values, the role of each value, the states a component can take, and where it all came from. None of that is in the pixels. A color sampled from an image can be shifted by compression and antialiasing. A font has to be guessed from its letter shapes. Hover, focus, and error states are absent unless someone captured them.
So turning an image gallery into agent input takes a separate step that estimates values from pixels, and the result is an estimate. Provenance matters for the same reason. A screenshot without its designer, date, or license cannot be traced back to anything that would confirm or correct the estimate.
Who is behind it
Ahmet Sulek is identified as CollectUI’s founder and curator and a Panda Network partner. Panda’s own material names Kubilay Kahveci, Serbay Arslanhan, Can Kutlu Kinay, and William Channer as part of the team that built it. The site is part of Panda Network, alongside related media properties.
Search snippets say Panda sends over 1.2M pageviews a month to Dribbble and has no investors. Both claims are UNVERIFIED because they were not fetched from a primary Panda page. CollectUI has no working /about page; the research got a 404 (not found) error. No homepage social links were found.
Format and how an agent consumes it
CollectUI has no structured design format. It offers no tokens, no DESIGN.md, no JavaScript Object Notation (JSON), no Cascading Style Sheets (CSS) variable exports, and no Really Simple Syndication (RSS) feed. It also has no application programming interface (API), command-line interface (CLI), Model Context Protocol (MCP) server, or browser extension. Its practical unit is an image in a category gallery.
So there is no agent-readable design document to quote. The closest accurate picture of the workflow is its navigation structure:
Home
Designers
Categories
Trending
/designs/card-ui-design-inspiration
/designs/user-profile-ui-design-inspiration
/designs/admin-panel-ui-design-inspirationAn agent cannot consume this as a specification. A person can inspect a screenshot, or a separate vision or extraction process could turn the image into something more formal. That second step is not a CollectUI feature. The research names a separate tool such as Design Extractor as the kind of layer required. Design Extractor reads rendered web pages rather than images, though, so a screenshot would still need a vision step first.
Pricing and access
No end-user subscription or paid browsing tier was observed. CollectUI is funded through ads and sponsorship. Its footer promotes sponsor slots and a self-serve advertising flow, with claims about dofollow backlinks, exposure to 1M+ designers, and 48-hour listing approval. These are footer claims, not independently audited figures.
No download gate or paid content tier was observed. Browsing is open.
What works
The category taxonomy is broad and easy to understand. It gives a person a quick way to find examples of a recurring interface pattern: cards, portfolios, widgets, profiles, or admin panels. A long-running collection is a useful mood-board starting point even without a machine-readable layer.
It also shows what an image gallery does well: qualitative visual context. A design engineer can see composition, density, hierarchy, and rhythm in a screenshot in a way that a bare token list cannot express.
What does not
For an agent, the lack of structure is decisive. There is no way to pull a token, component rule, provenance note, or source attribution into a project without a separate extraction step. The fetched category page showed no post dates or Dribbble-username attribution. That raises provenance and reuse questions.
The “updating daily” claim is UNVERIFIED, because fetched pages showed no dates to confirm freshness. Traffic figures from third-party summaries are likewise UNVERIFIED. Selling sponsorship and backlinks may also create a conflict of interest for curation, though the research found no specific instance of biased placement.
What we take from it
Velofy could propose a rich taxonomy for structured references. Card, widget, admin-panel, and portfolio categories are useful labels whether the payload is an image, a design file, or a component example. The proposal should pair the taxonomy with source addresses, dates, license context, extraction method, and machine-readable output.
It should not copy an image-only model for agent use. Velofy has not shipped a catalog or extractor. The gap is the proposal: keep the useful browsing structure while making the design evidence inspectable and attributable.